protocol
3D live fluorescence imaging of cellular dynamics
using Bessel beam plane illumination microscopy
Liang Gao1,2, Lin Shao1, Bi-Chang Chen1 & Eric Betzig1
1Janelia Farm Research Campus, Howard Hughes Medical Institute (HHMI), Ashburn, Virginia, USA. 2Present addresses: Department of
Chemistry, Stony Brook
University, Stony Brook, New York, USA; Department of Biochemistry and Cell Biology, Stony Brook University, Stony Brook, New York, USA. Correspondence should be
addressed to L.G. ([email protected]).
© 2014 Nature America, Inc. All rights reserved.
Published online 10 April 2014; doi:10.1038/nprot.2014.087
3D live imaging is important for a better understanding of biological processes, but it is challenging with current techniques such
as spinning-disk confocal microscopy. Bessel beam plane illumination microscopy allows high-speed 3D live fluorescence imaging
of living cellular and multicellular specimens with nearly isotropic spatial resolution, low photobleaching and low photodamage.
Unlike conventional fluorescence imaging techniques that usually have a unique operation mode, Bessel plane illumination has
several modes that offer different performance with different imaging metrics. To achieve optimal results from this technique, the
appropriate operation mode needs to be selected and the experimental setting must be optimized for the specific application and
associated sample properties. Here we explain the fundamental working principles of this technique, discuss the pros and cons
of each operational mode and show through examples how to optimize experimental parameters. We also describe the procedures
needed to construct, align and operate a Bessel beam plane illumination microscope by using our previously reported system as an
example, and we list the necessary equipment to build such a microscope. Assuming all components are readily available, it would
take a person skilled in optical instrumentation ~1 month to assemble and operate a microscope according to this protocol.
INTRODUCTION
Fluorescence microscopy is an essential technique used in
biological research to study the function and behavior of
proteins and organelles of interest. The applications of live
fluorescence imaging have expanded rapidly with the development of genetically encoded fluorescent probes and high-speed
high-photon-efficiency cameras based on either CCD or scientific
complementary metal-oxide semiconductor (sCMOS) technology,
which have provided subcellular spatial resolution and single
molecule sensitivity both in vitro and in vivo. However, in
practice, live fluorescence imaging is usually limited to two spatial
dimensions over a limited number of time points. Although 2D
live imaging is capable of providing basic dynamic information
of a living process, it is often incomplete and sometimes even misleading, for it only provides a partial representation of biological
processes that evolve in 3D volumes. Therefore, high-resolution
3D live imaging is required to obtain a complete and accurate
understanding of the biological processes being studied.
Challenges
3D live fluorescence imaging with subcellular spatial resolution
is challenging because it requires a good balance between spatial
resolution, optical sectioning, imaging speed, photobleaching,
photodamage and other factors. First, nearly isotropic 3D spatial
resolution, typically a few hundred nanometers, is necessary to
recognize subcellular structures in 3D environments. Second,
good optical sectioning capability is important for collecting
images with high signal-to-noise ratio (SNR) and low out-offocus background on thick samples or samples with densely
labeled structures. Third, the imaging speed has to be high enough
to capture the cellular and subcellular dynamics of interest, and
the speed must increase as the spatial resolution improves to avoid
blurred images. Fourth, the photobleaching and photodamage
induced by the imaging system must be low enough to allow
the specimen to be sampled densely enough and long enough to
reveal the biological process of interest while leaving the specimen
undisturbed so that the true physiological process is revealed.
It is worth mentioning that this problem represents a huge
challenge for many 3D live imaging techniques because at least
ten times and as many as 100 times or more 2D image planes must
be collected per 3D time point over typical cellular dimensions.
Finally, these issues do not exist independently; they are often in
opposition. For instance, higher spatial resolution almost always
results in lower imaging speed, increased photobleaching and
higher photodamage, and vice versa.
Conventional 3D live imaging techniques
Wide-field, confocal and two-photon (TP) fluorescence micro­
scopy techniques, based on an epi-illumination configuration
in which the same objective lens is used for excitation and
detection, have been essential for 3D live imaging. However,
a major problem of epi-illumination is that the excitation light
penetrates through the whole sample axially when only the sample
at the focal plane is in focus. The out-of-focus excitation produces
high fluorescence background, wastes the limited fluorescence
photon budget and introduces unnecessary photodamage. In
addition, epi-illumination leads to anisotropic spatial resolution
wherein the axial resolution is typically several times coarser than
the lateral resolution. This anisotropy often limits the observation of the specimen to the lateral direction only, even when the
imaging results are collected in 3D, because observations from
other directions are less informative.
Each epi-illumination method has additional, specific
limitations. Wide-field microscopy is not suited for 3D imaging
on specimens more than a few micrometers thick because it
is not able to reject fluorescence background created by this
out-of-focus excitation. Confocal microscopy gains its optical
sectioning capability by rejecting out-of-focus fluorescence by
using a single pinhole or a pinhole array, but the improvement is
nature protocols | VOL.9 NO.5 | 2014 | 1083
protocol
Figure 1 | Plane illumination microscopy with scanned Bessel beams.
(a,b) The light sheet of conventional plane illumination can be created by
a sheet-like nonscanning line focus (a) or by scanning a focused Gaussian
beam to produce a virtual light sheet (b). (c) Comparison of Gaussian beams
at 0.5 (top) and 0.15 (middle) excitation NA and a Bessel beam with 0.5
NAOD and 0.46 NAID (bottom). The Bessel beam has a central peak slightly
narrower than even the high-NA Gaussian beam, but it is at the same time
as long as the low-NA Gaussian beam. However, it also has strong side lobes,
which must be taken into account when it is used for plane illumination.
a
© 2014 Nature America, Inc. All rights reserved.
Selective plane illumination microscopy (SPIM)
Although the technique is a century old1, the rapid development
of SPIM over the last decade2–9 was driven by the need for highspeed, low-photobleaching and noninvasive 3D live imaging.
SPIM minimizes the out-of-focus excitation by using two objectives with optical axes orthogonal to each other separately for
excitation and detection. The sample is illuminated from the side
by sending a sheet of excitation light into the sample, so that the
illumination is confined near the focal plane of the orthogonal
detection objective (Fig. 1a). In comparison with epi-illumination
methods, SPIM has the following advantages. First, the confined
excitation provides optical sectioning automatically by producing
minimal out-of-focus fluorescence background. Second, the
method is as fast as conventional wide-field detection, as
fluorescence is imaged across the entire excitation plane
simultaneously. Third, if the sheet thickness is comparable to
or thinner than the depth of focus, the overall axial resolution
is improved. Finally, SPIM has much lower photobleaching
because only the fluorophores near the detection focal plane are
illuminated, and this confinement of excitation to one plane also
mitigates photodamage.
In SPIM, the light sheet can be a real one, existing simultaneously across the illumination plane (such as that produced by a
cylindrical lens)2,3, or a virtual one, created by serially scanning a
long Gaussian beam produced at a low numerical aperture across
the plane4–9. A real light sheet typically requires much lower peak
intensities so that it produces less photodamage, whereas a virtual light sheet gives more control over the size and the intensity
profile of the illuminated region.
Conventional SPIM has proven to be most successful for
the 3D live imaging of multicellular specimens with single-cell
resolution. The 3D spatial resolution of SPIM is determined
by both the excitation and detection numerical apertures
(NAexc and NAdet, respectively). The lateral resolution is the same
as the conventional diffraction limit of the wide-field microscopy,
lem
, where λem is the wavelength of the emitted fluorescence.
2 NA det
The axial resolution of
 2NA exc n(1 − cosq det ) 
+
 l

lem
 exc

1084 | VOL.9 NO.5 | 2014 | nature protocols
−1
is related to both
b
Detection objective
Excitation objective
Cylindrical
lens
Illumination light sheet
c
accompanied by slower imaging speed, faster photobleaching and
higher photodamage, thereby limiting the number of time points
for live imaging. TP fluorescence microscopy achieves optical sectioning and simultaneously avoids an out-of-focus background
through nonlinear absorption of the excitation, but the imaging
speed is still limited by point-scanned detection using the photomultiplier tube (PMT), and nonlinear mechanisms introduce
additional photodamage. It also has a limited multicolor imaging
capability. In practice, confocal and, more recently, spinning-disk
confocal microscopy have been the most common methods for
live 3D imaging of cellular-scale specimens.
Detection objective
Scanning
beam
Virtual illumination light sheet
Image plane
Pupil plane
10 µm
the thickness of the illumination light sheet,
l
lexc
2 NA exc
1 µm
, and the axial
em
resolution of detection objective, n(1− cos
q det ) , where λexc is the
excitation wavelength, n is the refractive index of the imaging
buffer, and θdet = sin−1(NAdet/n) is the half-angle of light collection in the imaging buffer. Clearly, the axial resolution becomes
higher with higher NAexc, corresponding to a thinner light sheet.
More importantly, the missing cone of the wide-field detection
optical transfer function (OTF) is filled in SPIM by the excitation
light sheet, which is the fundamental reason why SPIM has
optical sectioning capability.
To obtain uniform axial resolution and optical sectioning
capability, the light sheet thickness has to be uniform across
the entire imaging field. However, there is a trade-off between
the length of a Gaussian light sheet and its thickness owing to the
diffraction of light: a light sheet of minimum thickness of 2wo
varies in thickness by a factor of 2 over a distance defined by
the Rayleigh length, p w02 /lexc . For instance, a light sheet ~2 µm
thick is required to cover a ~30-µm field of view with a 488-nm
excitation wavelength. Therefore, Gaussian light sheets of
2–10 µm in thickness are commonly used. The axial resolution
with such light sheets remains dominated by the axial resolution
of the detection objective alone, and hence conventional SPIM
(without sample rotation)4–9 rarely reveals as much subcellular
detail in the axial direction as a confocal microscope at a high
NA. To improve the axial resolution to the point at which the
3D spatial resolution is nearly isotropic, and to maximize the
advantages of optical sectioning, thinner illumination light sheets
are required.
Bessel beam plane illumination microscopy
We introduced Bessel beams10 to overcome the trade-off between
the length and thickness of conventional Gaussian 2D beams and
1D light sheets. Below we discuss the three operational modes
of Bessel beam plane illumination microscopy (Table 1): Bessel
beam plane illumination (sheet-scan mode), TP Bessel beam
plane illumination (TP excitation sheet-scan mode), and Bessel
beam structured plane illumination (Bessel structured illumination microscopy mode).
protocol
Table 1 | Performance comparisons for different operational modes of Bessel beam plane illumination microscopy (scale bar, 1 µm).
Bessel SIM
TP Bessel beam
plane illumination
TP Bessel
SIM
x ~230 nm
x ~180 nm
x ~230 nm
x ~180 nm
y ~230 nm
y ~230 nm
y ~230 nm
y ~230 nm
z ~600 nm
z ~350 nm
z ~450 nm
z ~450 nm
Fair
Good
Excellent
Excellent
~40 planes/s
10–20 planes/s
~40 planes/s
10–20 planes/s
Multicolor imaging
Yes
Yes
No
No
Fluorophore brightness requirement
Low
Low
High
High
Photodamage
Low
Fair
Low
High
Operation modes
Bessel beam
plane illumination
Illumination pattern
(xz cross-section)
1
© 2014 Nature America, Inc. All rights reserved.
Typical spatial resolution
Optical sectioning capability
Typical imaging speed
Bessel beam plane illumination. An ideal Bessel beam, one of
the class of so-called ‘nondiffracting beams’, propagates indefinitely with no change in its cross-sectional intensity profile,
I(r)∝|Jo(αr)|2, where Jo(αr) is the zero-order Bessel function of
the first kind11. Thus, the central peak of this ideal beam is surrounded by an infinite series of concentric side lobes of decreasing peak intensity, but with equal integrated intensity in each
lobe. The ideal beam would exist at the front focal plane of a lens
illuminated at its back focal plane with an infinitesimally thin
annulus of light.
In practice, this is not possible. An experimental ‘Bessel beam’ is
actually a Bessel-Gauss beam created with annular illumination of
finite width. For a given NAexc, the thinner the annulus, the more
Bessel-like the beam becomes, the longer its axial extent (i.e.,
the beam length), and the more energy exists in the side lobes
(Fig. 1b). Conversely, the wider the annulus, the more Gaussian
the character of the beam, the shorter its axial extent, and the
smaller the side lobe energy. In either limit, the width of the
central peak, and hence the thickness of the central core of a
virtual light sheet made by sweeping the beam, is proportional
to
lexc
. However, for longer beams that cover larger fields of
2NA exc
view, the increased side lobe energy results in increased energy
in excitation tails on either side of this central core. Hence, a key
aspect of Bessel beam plane illumination is to use a beam that is
only as long as is necessary to cover the desired field of view and
no longer. Beyond this, other tricks, such as nonlinear excitation or structured illumination, are needed to mitigate the effects
of the reduced optical sectioning and increased out-of-focus
background and photobleaching resulting from the residual side
lobe energy.
TP Bessel beam plane illumination. Although substantial
integrated energy can exist in the side lobes, the intensity I of the
central peak of even an ideal Bessel beam is much greater than
that of any of the side lobes. By using a nonlinear process S∝In
with n ≥ 2, the signal S generated at the central peak dominates
and the contribution from the side lobes is minimal. For example, with TP excited fluorescence the signal can be constrained
to a thin beam of several hundred nanometers thick but tens of
micrometers long. Scanning this beam produces a virtual light
sheet having negligible out-of-focus fluorescence background.
At NAdet = 1.1, this TP mode offers a nearly isotropic 3D spatial
resolution of ~230 nm laterally and ~350 nm axially, as well as
excellent optical sectioning capability and fast imaging speed up
to hundreds of planes per second10. Its limitations are a relatively
high requirement on the brightness of fluorophores (compared
with linear excitation), a limited capability for live imaging with
multiple colors simultaneously and the possibility of additional
nonlinear photodamage mechanisms. In contrast, as with conventional TP imaging, the IR excitation light that is typically used
can penetrate further into multicellular organisms and tissues
than visible excitation. As a result, the TP mode is a good option
when imaging deeply into scattering and/or aberrating specimens,
and it can produce substantially thinner virtual light sheets than
Gaussian TP beams12 over a comparable field of view.
nature protocols | VOL.9 NO.5 | 2014 | 1085
protocol
Bessel beam structured plane illumination. The combination
of Bessel beam plane illumination with structured illumination
microscopy (SIM) provides another avenue to minimize the negative effects of the Bessel beam side lobes. In conventional widefield SIM, a periodic excitation pattern of period T is applied to
the sample, and a series of raw images are acquired as the pattern
is moved relative to the sample in N equal steps of T/N. One of
two different algorithms is then used to reconstruct an image of
the sample at the focal plane.
In optical sectioning SIM (OS-SIM) 13, the N raw
images In are combined to create the reconstructed image
© 2014 Nature America, Inc. All rights reserved.
I final =
N
∑ In exp(i2p n / N ) . The effect of this algorithm is to subtract
n =1
the contributions from the out-of-focus signal in the N raw images,
which is much more weakly modulated by the applied excitation
pattern than the in-focus signal.
In super-resolution SIM (SR-SIM)14, N raw images are again
collected, which generates a series of N equations that can be
combined with the N unknowns (given by the frequencyshifted copies of the sample structure encoded by the patterned
excitation) to produce a reconstructed image Ifinal beyond the
diffraction limit of the detection objective in the patterned
direction and, in 3D SIM15,16, along the detection axis as well.
To apply either SIM approach to Bessel beam plane illumination,
the beam is moved in discrete steps, rather than continuously,
across the excitation plane to create a virtual grating pattern for
SIM (Fig. 2). In contrast to wide-field OS-SIM, in which usually
only N = 3 images are acquired per plane (to minimize the time
per plane needed to achieve optical sectioning), Bessel OS-SIM
requires the full complement of N = H + 2 raw images per plane
for the reconstruction to be uniform across the entirety of Ifinal
(i.e., for the point spread function (PSF) after reconstruction
to be spatially invariant). Although this would suggest using a
pattern of small period T containing only one harmonic H to
minimize the number of raw images required, in thick or densely
labeled samples the contrast of the applied pattern relative to
the out-of-focus background due to the Bessel side lobes is often
too low in this limit to provide a successful reconstruction with
an acceptable SNR. Thus, in practice, the excitation period T is
tailored to each sample individually until it is just large enough
to provide sufficient contrast in the raw images and acceptable
SNR in Ifinal. For most samples, periods corresponding to N = 7
or 9 images per plane are sufficient. The primary advantages of
Bessel plane OS-SIM10 are its marked reduction in out-of-focus
background in the reconstructed image, and its computational
simplicity: it can be used to provide real-time feedback as the data
are acquired concerning sample viability, brightness and optimal
imaging conditions. One disadvantage of the technique is that the
nonlinearity inherent in the absolute value function used in the
OS-SIM reconstruction leads to image artifacts. Another is that
the OS-SIM algorithm uses the signal encoded by only two of
the N = 2H + 1 frequencies in the excitation OTF (Fig. 2b, right)
in the reconstruction, or only a small fraction of the total signal.
As a result, high excitation power and longer periods requiring
more images per plane are often needed to achieve acceptable
SNR, leading to low imaging speed, fast photobleaching and high
photodamage in the context of live imaging.
Bessel plane SR-SIM17 works much better for image reconstruction. The N raw images per plane are acquired identically
1086 | VOL.9 NO.5 | 2014 | nature protocols
to Bessel plane OS-SIM, whereas the reconstruction uses the
same algorithm as 3D wide-field SR-SIM. Unlike OS-SIM, this
algorithm uses the signal encoded by all N = 2H + 1 frequencies of the excitation OTF and places the frequency-shifted sample information contained in each in its appropriate place in an
expanded frequency space representation of the sample. This has
three beneficial consequences. First, much lower excitation power
and fewer raw images per plane are required to achieve an acceptable SNR. This leads to faster imaging, less photobleaching and
reduced phototoxicity. Second, the reconstructed image faithfully
represents the true sample structure at all length scales down to
the resolution limit of the method. Third, when applied to 3D
image stacks, this resolution limit is itself extended beyond the
classical diffraction limit in both the patterning direction (defined
as the x direction) and in the axial direction orthogonal to the
excitation plane (the z direction).
Quantitatively, the resolution limit of Bessel plane SR-SIM in
 2NA
H
−1
det +
=
the x direction is given by Lmin
 , and in the z
x
T
 ldet
min
direction by Lz
−1
 2NA exc n(1 − cosq det ) 
=
+
 . The resolution along
lem
 lexc

the beam propagation direction (the y axis) remains diffraction
lem
limited: Lmin
.
y =
2NA det
In our instrument, we chose NAdet = 1.1 to maximize the signal
collection efficiency within the constraints of using a waterdipping (i.e., no cover glass) objective with a long working
distance of 2 mm to accommodate large specimens. Owing to the
additional constraints that the excitation and detection objectives must be orthogonal, and that the excitation and detection
illumination cones cannot intersect, the maximum excitation
NA is limited to NAexc = 0.65. We had a custom achromatic longworking-distance water-dipping objective fabricated to reach
this limit in order to achieve the highest possible 3D resolution.
Specifically, with these objectives we can reach Lmin
nm (equivx = 150
min
alent to a wide-field microscope of NAdet = 1.75), L y = 240 nm,
= 270 nm by Bessel plane SR-SIM when imaging green
and Lmin
z
(e.g., GFP) fluorescence.
Implementing multiple scanning beams to reduce
photobleaching and phototoxicity
In fluorescence imaging, photobleaching and phototoxicity
depend not only on the total integrated dose of excitation photons
delivered to the sample but also on the instantaneous peak intensity, as nonlinear damage mechanisms can have an important role.
This poses a considerable problem for plane illumination microscopy with a virtual scanned light sheet, as the excitation energy
is then concentrated in a line that must sweep rapidly across the
illumination plane in the limited time allotted for collecting each
2D image of a 3D stack. Indeed, as the light sheet becomes thinner
and the imaging speed becomes higher (twin goals of Bessel beam
plane illumination), the problem rapidly becomes more acute. For
example, a Bessel beam of central peak width of 0.5 µm scanning
at 50 planes/s across a 50-µm field of view will illuminate any
given voxel for only 200 µs; therefore, the beam intensity must
be high enough to produce enough fluorescence photons within
this time to reach an acceptable SNR.
To mitigate this problem, we have used parallel arrays of J Bessel
beams. Each beam then needs to scan over only 1/J of the original
Real space
NAOD = 0.5,
NAID = 0.46
Arbitary periodic modulation
function M(x)
0
T
x (µm)
–5
–5
Frequency space
1/0.4
KZ (1/µm)
0
1/T
0
–1/0.4
–1/0.4
c
0
KX (1/µm)
0
KZ (1/µm)
1/0.8
0
1/0.4
–1/0.4
–1/0.4
0
KZ (1/µm)
0.5
–1/0.8
1/0.8
–1/0.3
0
OTF-1st harmonic
1/0.8
1/0.4
1/0.8
1/0.4
0.5
0
–1/0.4
1/0.4
0
KZ (1/µm)
3rd harmonic
1
Intensity
OTF-DC
0
0
–1/0.2
–1/0.8
0
KZ (1/µm)
SRSIM OTF XY
KY (1/µm)
Intensity
d –1/0.3
–1/0.8
20
1st harmonic
0
–1/0.4
1/0.4
0.5
0
–1/0.4
10
x (µm)
1/0.4
1
2nd harmonic
1
0
KX (1/µm)
Intensity
–1/0.8
–5
1/0.4
0.5
0
–1/0.4
0
Kx (1/µm)
DC harmonic
1
Intensity
Fourier transform of the modulation
~
function M(Kx)
A
KZ (1/µm)
b
5
0
x (µm)
P = 1.8 µm
5
z (µm)
z (µm)
5
KZ (1/µm)
a
KZ (1/µm)
Figure 2 | Theory of Bessel beam structured
illumination. (a) The structured illumination
pattern is the convolution of the excitation beam
intensity profile and a periodical modulation
function, M(x), in real space. (b) The Fourier
transform of the illumination pattern is the
product of the Fourier transform of the excitation
beam intensity profile and the Fourier transform

of the modulation function, M(K
x ), in frequency
space. The number of harmonics contained in
the illumination pattern is determined by the
excitation beam NAOD and the modulation period T.
For the sheet-scan mode, T needs to be less than
λ/(2NAOD). (c) The normalized intensity profile
of each modulation harmonic with excitation
wavelength 488 nm, NAOD = 0.5 and T = 1.8 µm.
The DC harmonic contains highest axial spatial
frequency information, but provides the worst
optical sectioning capability. All AC harmonics
provide better optical sectioning capability but
lower axial frequency limits compared with the
DC harmonic. (d) The corresponding OTF of each
modulation harmonic, which is the convolution of
the wide-field detection OTF with each modulation
harmonic in frequency space. A gamma value of
0.5 is applied to the DC harmonic OTF to make the
high-frequency contents visible. The Bessel
SR-SIM OTF is the sum of all OTFs at the
corresponding position of each OTF in frequency
space. The fundamental harmonic OTF gives the
best optical sectioning, and the highest frequency
harmonic gives the maximal resolution extension
in the x direction.
KZ (1/µm)
KZ (1/µm)
0
field of view in the same time, requiring
only 1/J as much power in each beam to
1/0.3
1/0.3
0
–1/0.2
1/0.2
0
–1/0.2
1/0.2
1/0.2
produce the same amount of signal per unit
0
1/0.1
–1/0.1
KX (1/µm)
KX (1/µm)
time across the whole plane. Although the
KX (1/µm)
OTF-2nd harmonic
OTF-3rd harmonic
total dose delivered to the sample is then
–1/0.3
–1/0.3
SRSIM-OTF XZ
–1/0.3
the same as with a single beam, the reduc0
0
tion in peak intensity often has a profound
0
effect on prolonging cell viability. An ancil1/0.3
1/0.3
1/0.3
lary advantage of Bessel plane SIM is that
–1/0.2
0
1/0.2
–1/0.2
0
1/0.2
–1/0.1
0
1/0.1
KX (1/µm)
KX (1/µm)
KX (1/µm)
each beam need be moved in only 1/J as
many discrete steps. As the response time
of the galvanometer mirror used for this purpose can be a rate- mode (Bessel beam plane illumination) offers high speed and
multicolor capability, but its optical sectioning capability is comlimiting factor, this also opens the door to faster Bessel SIM imaging.
promised by the Bessel beam side lobes. This makes it difficult
We create these linear beam arrays using a commercial Dammann
to extract information even close to the approximately 3× axial
grating-based diffractive optical element (DOE) that splits an
incoming beam into a fan of J outgoing beams of equal intensity resolution gain that it theoretically provides over conventional
wide-field microscopy. Nevertheless, it can still address important
at equal angular spacing. When placed next to an annular mask
conjugate to the rear pupil of the excitation objective, this creates biological questions18,19 when the side lobe background is not a
concern. The TP mode (TP Bessel beam plane illumination) offers
an array of J parallel Bessel-Gauss beams across a plane within the
sample. However, it is important that these beams be separated high-speed and near-isotropic 3D resolution, but with limited
sufficiently so that their side lobes do not overlap appreciably, multicolor capability20. It is the mode of choice for deep imaging,
owing to the better penetration of its IR excitation in scattering
because the DOE-generated beams do not have a well-defined
and/or aberrating tissues. The Bessel plane OS-SIM mode prophase relationship with one another, and they therefore produce
vides good optical sectioning, but it is very photon inefficient and
nonperiodic speckle patterns when they mutually interfere.
prone to artifacts. It is used primarily for real-time assessment
of Bessel SIM images for later off-line processing by the SR-SIM
Summary of modes and applications of Bessel beam plane
algorithm. This Bessel plane SR-SIM mode (Bessel beam strucillumination microscopy
Table 1 summarizes the various operational modes of Bessel tured illumination) provides the best resolution: slightly beyond
beam plane illumination microscopy. The linear swept sheet the diffraction limit in the x direction and nearly 3× beyond the
KZ (1/µm)
© 2014 Nature America, Inc. All rights reserved.
protocol
nature protocols | VOL.9 NO.5 | 2014 | 1087
protocol
© 2014 Nature America, Inc. All rights reserved.
wide-field limit in z. It also offers good photon efficiency, accurate
3D sample reconstruction and multicolor capability. Its primary
disadvantages are the added time required to take N images per
plane instead of one, and the noise introduced if the higher spatial
frequencies are reconstructed at greater amplitudes than those
that the SNR should allow. Finally, although it is not discussed
here, the TP mode and Bessel plane SR-SIM mode can be
combined to merge the high resolution of Bessel plane SR-SIM
with the good depth penetration of two-photon excitation17.
Together, these modes provide a range of solutions that balance
3D spatial resolution, optical sectioning, imaging speed, photobleaching and photodamage for different live fluorescence imaging
applications. In comparison with conventional 3D fluorescence
imaging techniques, especially confocal fluorescence microscopy,
Bessel beam plane illumination offers better performance by all
these metrics.
Limitations of Bessel beam plane illumination microscopy
Bessel beam plane illumination microscopy has several limitations. First, although Bessel plane SR-SIM can extend resolution
in the x and z directions slightly beyond the diffraction limit,
Bessel beam plane illumination is not capable of extending into
the sub-100 nm regime without being combined with other
super-resolution methods21–24 such as photoactivated localization microscopy (PALM), stimulated emission depletion (STED)
microscopy or nonlinear SIM. Second, the imaging depth is
usually limited to ~50 µm or less below the sample surface owing
to optical aberrations and light scattering induced by the sample.
The so-called ‘self-healing’ properties of Bessel beams do not
improve the imaging depth because the conical shell of light that
forms the Bessel beam is also aberrated or scattered within such
samples25. Thicker Bessel-Gauss beams (i.e., lower NAexc) with
less side lobe excitation (i.e., thicker annulus width) are often
preferred in these circumstances, although the axial resolution
is poorer. Third, the sample mounting conditions are different
from those to which most biologists are accustomed. Samples are
mounted on 5- or 8-mm-diameter cover glasses in order to fit in
the space between the two objectives. In addition, because these
are water-dipping lenses, the specimen cannot be sandwiched
between cover slips: nothing can interfere in the half-space above
the cover slip. Fourth, the cover slip must be placed directly into
the imaging chamber clipped in a custom holder instead of into
a conventional Petri dish. In its current incarnation, the chamber
contains 10–35 ml of imaging buffer, which is advantageous for
maintaining temperature, pH and dissolved CO2 levels, but it can
be disadvantageous when one wishes to add expensive drugs in
perturbation experiments. Finally, every few weeks this chamber
must be disassembled and thoroughly cleaned to prevent
contamination, a process that requires several hours of skilled
realignment of the objectives and associated optics.
Spatial resolution and optical sectioning capability of Bessel
beam plane illumination microscopy
Axial resolution and optical sectioning. Two concepts, axial
resolution and optical sectioning capability, are particularly
important in 3D fluorescence imaging, and yet they are often
confused with each other. The axial resolution defines the smallest
axial separation at which two features can be resolved under
1088 | VOL.9 NO.5 | 2014 | nature protocols
conditions of negligible noise, whereas the optical sectioning
capability is a measure of the ability of a microscope to reject
out-of-focus fluorescence background. They are, however, related,
as increasing fluorescence background from poorer optical sectioning will set a practical limit on the axial resolution that is
coarser than the theoretical one.
These concepts can perhaps be better understood in frequency
space. A microscope image is the convolution of the microscope
PSF and the fluorescence emission pattern of the object under
investigation, D(r) = H(r)E(r), where D(r) is the recorded image,
H(r) is the PSF of the microscope and E(r) is the emission pattern.
 (k ) = H
 (k )E (k ),
By the convolution theorem, this is equivalent to: D
 (k ) , H
 (k ) , and E (k ) are the Fourier transforms of D(r),
where D
 (k ) defines the microscope’s OTF.
H(r) and E(r), respectively. H
It is nonzero only for a contiguous region of k space centered
at k = 0 and, within this region, it generally decreases with
increasing distance from the origin. Thus, the microscope acts as
a low-pass filter, and the recorded image that we see is a blurred
copy of emission pattern of the sample that we wish to know.
Broad features corresponding to low k are more strongly represented in the image than fine features corresponding to high k,
and no features corresponding to k values outside the nonzero
portion of the OTF will appear at all. In other words, the nonzero
region of the OTF defines the pass-band of the microscope and
the boundaries of this region, called the support of the OTF,
define the best possible spatial resolution theoretically obtainable in any given direction.
Seen in this light, the optical sectioning capability of a microscope is a reflection of how quickly its OTF rolls off to zero in the
axial direction: even with the same ultimate limit of support and
hence the same theoretical axial resolution, an OTF that maintains significant amplitude at high kz will outperform one with
lower amplitudes at the same kz, as the information in the image
will then be weighted more highly to the in-focus fluorescence
emitted closer to the focal plane than to the out-of-focus fluorescence from further away. In particular, in live fluorescence
imaging, the number of photons in the image is always limited
by photobleaching and phototoxicity, so the attenuated highfrequency signal is often submerged in noise from the out-offocus background, thereby making the theoretical resolution
unachievable in practice. Thus, the optical sectioning capability
of a microscope is at least as important as its theoretical axial
resolution limit.
Axial resolution and optical sectioning capability of Bessel beam
plane illumination. The axial resolution and optical sectioning
capability of Bessel beam plane illumination is determined by the
convolution of the traditional wide-field detection OTF with the
excitation OTF defined by the Bessel beam and the microscope’s
mode of operation. The detection OTF has a ‘missing cone’ of axial
information, which, by itself, leads to poor optical sectioning and
a substantial out-of-focus background. However, the OTF of the
excitation plane fills in this missing cone, reducing background
and often improving the practical limit of axial resolution.
In real space, the Bessel plane excitation function I(r) is the
convolution of the fluorescence emission function created by a
single Bessel-Gauss excitation beam B(r) and a function M(x)
that describes how the intensity of this beam is modulated as
protocol
it sweeps in the x direction to create a virtual excitation plane,
 (k ). For long
I(r) = B(r)M(x), or, in Fourier space: I (k ) = B(k )M
x
Bessel-Gauss beams, we can assume that the illumination is uniform along the propagation direction y over cellular-sized fields:
B(r) ≈ B(x, z). In this case, for each of the modes described above
and summarized in Table 1, we have:
Linear swept sheet: I(r) = B (x,z)Io, I (k ) = I o B (0, kz )d (kx )
Linear Bessel plane SIM:
I (r) = B(x , z ) ⊗ I o
I (k ) = I
o
H
∑
m = −H
H
∑
m = −H
Cm exp(i2p mx / T + iϕm )
 (m / T , k )δ(k − m / T )exp(iϕ )
Cm B
z
x
m

Two-photon swept sheet: I(r) = B2(x,z)Io, I (k ) = I o B2 (0, kz )d (kx )
© 2014 Nature America, Inc. All rights reserved.
Two-photon Bessel plane SIM:
I (r) = B2 (x , z ) ⊗ I o
I (k ) = I
o
H
∑
m = −H
H
∑
m = −H
Cm exp(i2p mx / T + iϕm )

Cm B2 (m / T , kz )d (kx − m / T )exp(iϕm )
where, as above, T is the period of the SIM grating pattern, and
H = floor [T/(0.5λexc/NAexc)] is the number of harmonics in
this pattern when the pattern is created by moving the beam in
discrete steps of length T. Note that in all of these expressions,
I (k ) , is of the form:
I (k ) = I
o
H
∑
m = −H

Cm B j (m / T , kz )d (kx − m / T )exp(iϕm )
where j = 1 or 2 (linear or two photon, respectively) and m = 0 or
m >0 (swept or SIM modes, respectively). In addition, the Bessel
plane excitation function I(r) creates a fluorescence emission pattern E(r) in a sample with a fluorophore distribution S(r): E(r) =
S(r)I(r). Thus, referring back to the beginning of this section:
 (k ) = H
 (k ) ⊗ S(k ) ⊗ I (k )
D


= Io
H
j

 (k ) ⊗ S(k ) ⊗  B
Cm H
(m / T , kz )d (kx − m / T )   exp(iϕm )




m = −H
∑
However, as the excitation plane remains coincident with the
detection focal plane throughout the capture of the 3D stack, it
has been shown that this reduces to14:
 (k ) = I
D
o
≡ Io
H
∑
m = −H
H
∑
m = −H
j
 (k ) ⊗ B
Cm  H
(m / T , kz )  S(k − (m / T )
e x )exp(iϕm )


CmOm (k )S(k − (m / T )
e x )exp(iϕm )
In physical terms, this expression states that the recorded image
is a sum of 2H + 1 copies of the sample structure, shifted in frequency space in successive increments of ∆kx = 1/T, and each
weighted by the strength Cm of the mth band of the periodic
j
 (k ) ⊗ B
(m / T , kz ),
excitation as well as by an effective OTF, Om (k ) = H
which serves as a filtering function of finite bandwidth.
Examples of Om(k) for different m and j = 1 (linear excitation) are shown in Figure 2d. The OTF corresponding to
the DC harmonic is equivalent to the overall OTF of the swept
sheet mode (i.e., where H = 0, so only the DC harmonic remains).
It gives the highest theoretical axial resolution, but has very
poor sectioning capability, as evidenced by the strong peak at
k = 0. This is a reflection of the strong out-of-focus background
from the side lobes of the swept Bessel beam. Moving outward
from the DC band, successive OTFs Om(k) corresponding
to bands at higher and higher |kx| provide less and less axial
resolution but more and more resolution in the x direction.
Together, in the above sum, they are capable of providing
substantial extension of resolution in the xz plane beyond the
wide-field diffraction limit of the detection objective alone
(Fig. 2d, right). To do so, however, the SNR of the collected images
and the coefficients Cm must be sufficiently large to detect their
contributions over the much stronger DC band, which has the
dominant peak at k = 0.
There are three ways to increase the strengths of the AC weighting
terms CmOm(k) relative to the DC term. The first is to write a
stepped pattern of longer period T, having more harmonics H.
It is realized by parking the Bessel beam at discrete positions of
fixed spacing T during scanning, while turning off the beam at
all other points in between. This tends to spread the spectral
energy out from the DC band across the additional harmonics.
In addition, the added harmonics at lower spatial frequencies
tend to have higher coefficients Cm than those closer to the
diffraction limit. Of course, more harmonics require more images
D(r) to reconstruct S(r); typically 5–7 raw images are required
per plane, which slows the imaging speed proportionally relative
to the sheet-scan mode. In addition, the information encoded by
the weak higher harmonics (which provide the greatest resolution extension in the x direction) may not be recoverable. Thus,
another approach is to sweep the beam continuously, but to modulate its amplitude with the acousto-optical tunable filter (AOTF)
in a sine pattern of period T. This generates only H = 1 harmonic,
with relatively large Cm for m = ±1, as the associated bands are
at low |kx|. It requires only three images D(r) to reconstruct S(r),
so that the imaging speed is only three times slower than that
of the sheet-scan mode. It also induces less photodamage
compared with the stepped mode, yet it extends the resolution
in the z direction just as much, as both approaches exploit the
high |kz| extent of the first harmonic. However, this sine modulation mode does not provide much resolution improvement in
the x resolution as the first harmonic is restricted to low |kx|. The
final approach is to reduce the spectral energy in the DC term,
particularly in the peak near k = 0. This can be done by choosing
a Bessel-Gauss beam having less side lobe excitation, and hence
an out-of-focus background contributing to this peak. Of course,
this means choosing a shorter beam and reducing the field of
view in the y direction.
nature protocols | VOL.9 NO.5 | 2014 | 1089
protocol
© 2014 Nature America, Inc. All rights reserved.
MATERIALS
REAGENTS
• Alexa Fluor 488, hydrazide, sodium salt (Invitrogen, cat. no. A10436),
see Reagent Setup
• Alexa Fluor 568, hydrazide, sodium salt (Invitrogen, cat. no. A10437),
see Reagent Setup
• Alexa Fluor 647 cadaverine, disodium salt (Invitrogen, cat. no. A30679),
see Reagent Setup
• FluoSpheres carboxylate-modified microspheres, 0.1 µm, yellow-green
fluorescent (Invitrogen, cat. no. F8803) see Equipment Setup
• FluoSpheres carboxylate-modified microspheres, 0.1 µm, red fluorescent
(Invitrogen, cat. no. F8801) see Equipment Setup
• FluoSpheres carboxylate-modified microspheres, 0.2 µm, dark red
fluorescent (Invitrogen, cat. no. F8807) see Equipment Setup
• Dulbecco’s PBS, 1× PBS (Cellgro, cat. no. 21-031)
• Poly-d-lysine hydrobromide (Sigma-Aldrich, cat. no. P0899)
• Cell-Tak cell and tissue adhesive (BD Biosciences, cat. no. 354240)
• Hydrogen peroxide, 50% (wt/vol) (H2O2) ! CAUTION It is corrosive.
• Ammonium hydroxide, 30% (wt/vol) (NH4OH) ! CAUTION It is corrosive.
EQUIPMENT
 CRITICAL The opto-mechanical model for our Bessel beam plane
illumination microscope and the associated mechanical drawings for
custom components are available from HHMI after execution of a
nondisclosure agreement (NDA)
 CRITICAL We have listed the equipment used to build our microscope.
However, unless otherwise specified, most items can be replaced with
products from other vendors that give comparable performance.
• Optical table (TMC, optical tabletop, cat. no. 784-455-02R, optical table leg,
cat. no. 14-417-35)
• Optical breadboard (ThorLabs, cat. no. PBH11106)
Laser sources for linear excitation and TP excitation
• 405-nm laser (Coherent, CUBE, 405-100C)
• 488-nm laser (Coherent, Sapphire LP, 200 mW)
• 561-nm laser (Coherent, Sapphire LP, 200 mW)
• 640-nm laser (Coherent, CUBE 640-100C)
• Ultrafast Ti-sapphire laser (Coherent, Chameleon Ultra II Ti-Sapphire laser)
Laser intensity modulation devices
• AOTF, for 405 nm~650 nm continuous-wave (CW) lasers (AA Opto-Electronic,
cat. no. AOTFnC-400.650-TN)
• Pockels cell (Conoptics, cat. no. M350-80LA-02, for the Ultrafast laser)
Laser, objective and sample-scanning devices
• Scanning galvanometer mirrors, 3 mm square, protected silver (GM-Z,
GM-X, Cambridge Technology, cat. no. 6215HP-1HB)
• Objective scan piezo, 100-µm range (Physik Instrumente,
cat. no. P-726.1CD PIFOC)
• Objective scan piezo controller (Physik Instrumente, cat. no. E665)
• Sample scan piezo, 100-µm range (Physik Instrumente,
cat. no. P-621.1CD)
• Sample scan piezo controller (Physik Instrumente, cat. no. E625)
Objectives
• Excitation objective (EO, Nikon CFI Apo 40XW near-infrared (NIR),
0.8 NA, 3.5 mm working distance (WD) or Special Optics, custom objective,
cat. no. HH-PO159319, water dipping, 0.65 NA, 3.74 mm WD)
• Detection objective (DO, Nikon CFI Apo 40XW NIR, 0.8 NA, 3.5 mm WD
or Nikon CFI Apo LWD 25XW, 1.1 NA, 2 mm WD)
• Wide-field objective (WO, Olympus, UMPLFLN 20XW, 0.5 NA,
3.5 mm WD)
Detection camera
• Electron multiplying (EM)-CCD camera (Andor Technology, iXon3 885,
high quantum efficiency)
• sCMOS camera (Hamamatsu, ORCA Flash 4.0, higher speed, larger filed
of view)
• Alignment inspection camera, wide-field camera (AVT Guppy F-146
Firewire, Edmund Optics, cat. no. 59-731)
Laser shaping devices
• DOE (Holo/OR, cat. no. MS-107-Q-Y-A, to create seven Bessel beams
simultaneously)
• Custom aluminum-coated quartz annulus mask (AM, Photo Sciences, to
create an annular laser beam)
• Axicon, 0.5° (AX, Thorlabs, cat. no. AX2505-B, to create an annular focus)
1090 | VOL.9 NO.5 | 2014 | nature protocols
System control devices
• Computer, >48 GB RAM and >1 TB hard drive (Supermicro,
cat. no. SYS-7046GT-TRFP. Recommended configuration:
2 Xeon 5680 CPU)
• Rack-mount LCD monitor (Crystal-Image Technologies,
cat. no. RMPW-161-19))
• Signal generation board (National Instruments, cat. no. PCIe-7852R)
• Scaling amplifier mainframe (Stanford Research Systems, cat. no. SIM900)
• Scaling amplifier (Stanford Research Systems, cat. no. SIM983)
• Control software written in Labview 2013 (available from HHMI upon
execution of an NDA). National Instruments (NI) Labview, the NI Vision
Development Mmodule and the NI field-programmable gate array (FPGA)
module are required to run the software
Filters and dichroic
• Neutral density filter wheel (Thorlabs, cat. no. NDM2)
• Lasermux laser combining filter (LM, Semrock, cat. no. LM01-427-25)
• Lasermux laser combining filter (LM, Semrock, cat. no. LM01-503-25)
• Lasermux laser combining filter (LM, Semrock, cat. no. LM01-613-25)
• Short-pass dichroic beam splitter (DC1, Semrock, cat. no. FF670-SDi01-25x36)
• Multiband laser-flat dichroic beamsplitter (DC2, Semrock,
cat. no. Di01-R405/488/543/635-25×36)
• Band-pass filter (Semrock, cat. no. FF01-525/50-25), to be used as the
emission filter in the detection path for fluorescent probes with emission
wavelength at ~525 nm. For instance, GFP, mEmerald and Alexa Fluor 488
• Band-pass filter (Semrock, cat. no. FF01-600/37-25), to be used as the
emission filter in the detection path for fluorescent probes with emission
wavelength at ~600 nm. For instance, RFP, tdTomato, mCherry and Alexa
Fluor 568
• Multiband band-pass filter (Semrock, cat. no. FF01-523/610-25), to be used
as the emission filter for dual-color imaging by using fluorescence probes
with emission wavelength at ~525 nm and ~600 nm
• Multiband band-pass filter (Semrock, cat. no. FF01-440/521/607/700-25), to
be used as the emission filter in the wide-field detection path
Other optical components
• Achromatic lenses (focal lengths should be selected on the basis of the optical system design)
• Optical mirrors (protected silver)
Sample and sample mounting
• Sample. Samples should be prepared, fixed and labeled using the same
methods used for regular fluorescence microscopy (wide-field, confocal)
• Sample coverslip (Warner Instruments, cat. no. CS-5R/CS-8R)
• Sample mounting. Cellular specimens should be cultured directly on
either 5-mm- or 8-mm-diameter coverslips, which can then be clipped
onto the top surface of the specimen holder and mounted in the chamber
with the sample side facing the excitation and detection objectives.
Multicellular organisms, such as Caenorhabiditis elegans or Drosophila
embryos, can be attached to the cover slip with poly-d-lysine, tissue
adhesive or double-sided tape. Larger samples can be mounted with
agarose gel according to published protocols4
REAGENT SETUP
Alexa Fluor solutions Prepare separate 0.5 µg/ml solutions of Alexa Fluor
488 hydrazide sodium salt in H2O for system alignment using the 488-nm
CW laser and the Chameleon ultrafast laser at 920 nm. Prepare Alexa Fluor
568 hydrazide sodium salt solution and Alexa Fluor 647 cadaverine, disodium
salt solution with the same concentrations for the alignment of the 568-nm
CW laser and the 647-nm laser alignment, respectively.
EQUIPMENT SETUP
Preparation of cleaned cover slips Prepare a 150 ml mixture solution of
hydrogen peroxide (H2O2, 50%), ammonium hydroxide (NH4OH, 30%) and
H2O in a ratio of 1:1:5 by volume. Soak the coverslips in the solution for 12 h.
Rinse the coverslips with H2O and sterilize by flaming them with methanol.
The cleaned coverslips can be stored in sterile containers for up to 1 week.
! CAUTION These are corrosive chemicals; coverslips should be prepared in a
chemical hood.
Preparation of fluorescent bead sample for microscope alignment Dilute
fluorescent microsphere solution of the appropriate color for the laser
wavelength of interest 106× in H2O. Add ~30 µl of a solution of 0.1 mg/ml
poly-d lysine in water on top of a cleaned 5-mm coverslip and leave it to dry
for 15 min. Rinse the coated coverslip with water. Add ~30 µl of the diluted
protocol
561 nm
635 nm
635 nm
488 nm
488 nm
561 nm
405 nm
AOTF
© 2014 Nature America, Inc. All rights reserved.
AOTF
405 nm
Figure 3 | Bessel SPIM microscope schematic
a
c
Camera
Linear excitation
diagram. (a) CW laser sources at 405 nm, 488 nm,
Specimen chamber
Ultrafast laser
PC
DO
561 nm and 635 nm are used for linear
excitation. The four collimated laser beams are
AM
L4
TL1 L3
each expanded to a 1.5-mm diameter and then
GM-X
AX
L7
combined by using LaserMux dichroic mirrors
DOE
L2
L5
Specimen
(LM) into a single beam that is sent to the AOTF.
L8
chamber
EO
RND
Scanning galvo mirrors, GM-Z and GM-X; annulus
DO
z
L1
L6 EO
FM1
mask (AM); the rear focal plane of excitation
Coverslip
y
WO
x
FM2
GM-Z DC1
objective, EO; and a rear pupil inspection camera
LM
LM
LM
M
WO
are conjugated with each other in sequence
DC2
Inspection
Wide-field
through relay lens L1 and L2 (45-mm and 90-mm
Objective scan
Sample scan
TL2
camera
camera
focal length), L3 and L4 (50-mm and 150-mm
focal length), L5 and L6 (125-mm and 125-mm
b TP excitation
d
Camera
focal length). DOE can be placed immediately
Annulus mask
in front of mask AM to generate an array of
Ultrafast laser
PC
parallel Bessel beams. Detection objective, DO,
TL1 L3
AM
L4
is placed orthogonal to excitation objective EO.
AX
GM-X
The emitted fluorescence is collected through DO
L7
L2
FP1
and imaged onto the detection camera with tube
L5
L8
lens TL1 (500-mm focal length). Flip mirror FM1
RND
DO
L1
L6 EO
DC1
is used to divert the multicolor excitation to a
GM-Z
separate wide-field imaging path that serves as
FM2
L9 FM1
M
LM
LM
LM
WO
a low-magnification view finder. Flip mirror FM2
DC2
Inspection
is used to direct the annular excitation beam to
Wide-field
TL2
camera
camera
the inspection camera. (b) A Chameleon ultrafast
laser is used for TP excitation. A Pockels cell is
used for laser intensity modulation, and the NIR beam is collimated and expanded to a 10-mm diameter. Axicon AX (0.5° cone angle) and lens L7 (100-mm
focal length) are used to create an annular focus at focal plane FP1. This is then conjugated to GM-Z and all subsequent conjugation planes through relay lens
L8 L9 (50-mm and 150-mm focal length). (c) Schematic of the specimen chamber, which consists of the chamber itself, excitation objective (EO), detection
objective (DO), the specimen holder, and the wide-field view-finding objective. 3D images of the sample can be collected either by scanning the light sheet
and DO in synchronization with one another, or by scanning the sample through the light sheet. (d) Mask AM contains an array of annuli array of differing
outer and inner diameters to create Bessel beams of differing lengths and thicknesses. (This model and the associated mechanical drawings for custom
components are available from HHMI after execution of a NDA.)
fluorescent bead solution to the coverslip and leave it to dry. Rinse the
coverslip again with water and the sample is ready for immediate use. The
sample can be stored in the dark at room temperature (20–25 °C) and reused
until the fluorescent microspheres are bleached.
Bessel beam plane illumination microscope design The schematic
diagram of our Bessel beam plane illumination microscope is shown in
Figure 3, with the corresponding 3D opto-mechanical model shown
in Supplementary Figure 1. This model and the associated mechanical
drawings for custom components are available from HHMI after
execution of a NDA. Most standard optical elements are mounted in
30-mm cage system components available from Thorlabs. Beam steering
mirrors are all of λ/10 flatness. Those shared by both the ultrafast and CW
lasers have protected silver coatings, whereas those used by the CW lasers
alone have broadband dielectric coatings.
CW laser sources with excitation wavelengths at 405, 488, 561 and
635 nm are used for single-color and multicolor imaging via linear
excitation. Each laser beam is collimated and expanded to a 1.5-mm
beam diameter (1/e2) and then combined in a single co-linear beam before
entering an AOTF (Supplementary Fig. 2a). The AOTF is used to select
the currently active wavelengths and to control their amplitudes.
The output beam from the AOTF is sent to beam-scanning components
(Supplementary Fig. 2b) that consist of two 3-mm-square scanning
galvanometer (galvo) mirrors. The first, GM-Z, scans the virtual light sheet
along the detection axis and is conjugated at 2× magnification to the second,
GM-X, which scans the beam laterally to create the virtual light sheet. This
second galvo is then conjugated at 3× magnification to an opaque mask
AM (Fig. 3d) containing transmissive annuli of different maximum and
minimum diameters defining different maximum and minimum NA for the
eventual Bessel beam. This annulus mask is mounted on a 3D translational
stage to allow accurate positioning and switching to different annular
patterns. The annular beam emerging from mask AM is then conjugated to
the rear focal plane of excitation objective EO.
If desired, a DOE can be placed in front of mask AM to generate an array
of parallel excitation Bessel beams within the sample. However, because the
individual beams from the DOE do not have similar phases, separation
between the beams within the sample should be larger than the effective
range of interaction between the side lobes of adjacent beams.
Detection objective DO, mounted on a detection objective piezo, is placed
orthogonally to excitation objective EO (Supplementary Figs. 3a and 4).
In the objective-scanning implementation, a 3D image stack is acquired
by synchronizing the motion of the piezo to galvo GM-Z (black arrows,
Fig. 3c) so that the excitation light sheet always overlaps with the detection
focal plane. The emitted fluorescence is collected through the detection
objective DO and imaged on the detection camera with a tube lens. The
effective pixel size at the camera should be slightly smaller than the Nyquist
limit of λdet/(4NAdet) to achieve the full theoretical resolution of the system.
In the TP modes, an ultrafast Ti-sapphire laser of tunable wavelength is
used as the excitation source (Fig. 3b). The NIR laser beam passes through
a Pockels cell for laser intensity modulation. To maximize the transmission
laser power of the NIR laser beam through mask AM, NIR achromatic lens
L7 and axicon AX are paired to create an annular ring of excitation at the
front focal plane of lens L7 (ref. 26). This plane is conjugated to galvo GM-Z,
and hence mask AM as well, at a magnification such that the dimensions
of the excitation ring at AM match those of the chosen annulus. With this
approach, more than 60% of the laser power can be transmitted through the
annulus. A similar achromat/axicon subsystem can be placed between the
AOTF and galvo GM-Z when more excitation energy is required for Bessel
plane illumination with linear excitation.
The specimen chamber subassembly is shown schematically in Figure 3c
and from the 3D model in Supplementary Figure 3a. The chamber itself
(Supplementary Fig. 3b) has a vertical slot that is filled with the imaging
buffer into which the specimen holder is inserted. There are also side ports
for insertion of the excitation and detection objectives, as well as a third
objective to provide an epifluorescence view of the sample to locate the
nature protocols | VOL.9 NO.5 | 2014 | 1091
protocol
© 2014 Nature America, Inc. All rights reserved.
Figure 4 | Control signal sequences for Bessel
beam plane illumination. (a) General control
signal sequences for 3D imaging. Control
signals for the AOTF/Pockels cell, galvos GM-X
and GM-Z, the detection-objective piezo and
the sample-scan piezo are triggered by the
imaging camera. (b) Control signal sequences
for the sheet-scan mode at each image plane.
(c) Control signal sequences for the SIM mode
with sinusoidal modulation at each image plane.
(d) Control signal sequences for the SIM mode
with comb modulation at each image plane.
Phase shifting of the illumination pattern is
realized by changing the initial position of the
excitation beam.
a
Bessel SPIM signal sequence
Camera exposure on/off
X galvano mirror
AOTF/Pockels cell
Z galvano mirror and
objective piezo
/sample piezo
b
Bessel SPIM sheet scan
Camera exposure on/off
X galvano mirror
AOTF/Pockels cell
region of interest during imaging. A piece of
0.1-mm-thick silicone plastic film is stretched
over the sides of the excitation, detection and
c Bessel SIM comb modulation
epi-illumination objectives and is clamped to
Camera exposure on/off
the sides of the sample chamber to contain the
medium in the sample chamber while still
X galvano mirror
permitting the objectives to move with minimal
Initial offset:
P/3
2P/3
P
resistance. A serpentine circulation channel is cut
AOTF/Pockels
cell
in the specimen chamber to allow heated water to
flow in a separate circuit through the body of the
dBessel SIM sinusoidal modulation
chamber to keep the medium and the specimen
at the desired temperature (e.g., 37 °C for live
Camera exposure on/off
mammalian cell imaging).
All three objectives (EO, DO and WO) must
X galvano mirror
be long-working-distance, water-immersion
objectives. Owing to the physical size constraints
Initial offset:
P/(2N + 1)
2P/(2N + 1)
P
between the excitation and detection objectives,
AOTF/Pockels cell
the sum of the half-angles of their illumination
cones must be less than 90°, and they must not
mechanically intersect when orthogonal to
maximum recording rate of the Orca-Flash 4.0 sCMOS camera does
one another and placed at a common focus (Supplementary Fig. 4).
not overload the typical ~100 MB/s write speed of a single hard drive.
We have used two pairs of objectives (EO and DO) that satisfy these conAn external RAID array also offers considerably greater storage space than a
straints. The first combination uses identical Nikon CFI Apo 40XW (0.8 NA,
single drive, which can be easily filled in a couple of days of experiments.
3.5 mm WD) objectives as both the EO and DO, and the other uses a custom
A NI PCIe-7852R multifunctional reconfigurable I/O board triggered by
water immersion objective designed and built by Special Optics (0.65 NA,
the imaging camera is used to generate control signal sequences through the
3.74 mm WD) as the EO and a Nikon CFI Apo LWD 25XW (1.1 NA, 2 mm
connection box CA-1000 to synchronize the AOTF or Pockels cell, galvos and
WD) as the DO. The 1.1-NA Nikon objective also has a correction collar
objective-or sample-scanning piezos with image collection. Analog control
that can be used to correct the spherical aberration induced by the refractive
signals for the galvos and piezos are further conditioned by individual scaling
index differences of different imaging buffers and at different temperatures.
amplifiers to better match the 16-bit resolution output of the FPGA card
Its axial resolution is also better matched to the transverse resolution of the
to the input signal range of each device. Customized control software was
0.65-NA excitation lens, and its light collection ability is superior to the
developed in Labview 2012 by Coleman Technologies, and it includes
0.8-NA configuration. Thus, the second objective combination is preferred.
routines for microscope calibration and operation under different imaging
The specimen holder (Supplementary Fig. 4) is mounted on a
modes, as well as tools for reviewing the collected data in situ. This software
sample-scanning piezo that is mounted in turn on an xyz translational
is also available from HHMI upon execution of an NDA. Typical control
stage designed to position the sample region of interest to the volumetric
signal sequences for the various operational modes of Bessel beam plane
region swept out by the Bessel beam. In the sample-scanning
illumination are shown in Figure 4.
implementation, a 3D image stack is acquired by translating this piezo
Scanning mode The volumetric image of the sample can be obtained either
(green arrows, Fig. 3c), whereas the Bessel beam plane and detection
by objective scan or sample scan. In objective-scan mode, both the detection
objective remain stationary.
objective and the excitation light sheet are scanned across the sample simultaThe microscope is controlled with a Supermicro SYS-7046GT-TRFP
neously while a stack of images are collected sequentially during scanning at
computer equipped with dual hexa-core processors (Xeon X5680) and
different axial positions. Objective-scan mode is suited for imaging samples
96 GB of RAM. The computer backplane contains seven PCIe slots and two
with similar dimensions in all directions, e.g., C. elegans embryo. In contrast,
PCI slots, providing enough space for the necessary control cards. Either an
sample scanning can be used to image thin samples, as the excitation beam
Andor IXon 885 EM-CCD camera or a Hamamatsu Orca-Flash 4.0 sCMOS
can be much shorter than that in the objective-scan mode. This substantially
camera is used for image detection. Collected images are written directly to
reduces the excitation beam side lobes. In sample-scan mode, the excitation
a dual-disk array consisting of two SRS 7,200-r.p.m., 1-TB disks during
light sheet and the detection objective are fixed while the sample is scanned
imaging. Four-dimensional data stacks smaller than the RAM buffer can be
across the light sheet for 3D imaging (Fig. 3c). The data analysis is slightly
stored in RAM and then downloaded to the drives later. Larger data arrays
different for raw data collected by the two operation modes.
are best written directly to an external RAID array so that the ~800 MB/s
1092 | VOL.9 NO.5 | 2014 | nature protocols
protocol
© 2014 Nature America, Inc. All rights reserved.
PROCEDURE
Collimate and combine CW laser beams ● TIMING ~10 h
 CRITICAL The alignment of the Bessel beam plane illumination microscope should comply with general rules and laser
safety regulations for optical system alignment. Special care should be taken while aligning the NIR beam for TP excitation.
 CRITICAL The system should be aligned at ambient temperature, which should fall in the range of 20–25 °C. The actual
temperature variation should be less than 1 °C; variation in temperature causes substantial focal length drifting of the
microscope objectives.
 CRITICAL Initial alignment (Steps 1–23) needs to be performed only for the initial construction of the microscope;
commence subsequent imaging experiments at Step 24.
1| Collimate each beam by using its specific beam-expanding components so that all beams have a common beam
diameter. Combine all CW laser beams into a single coincident beam before the AOTF. Align the rest of the microscope
by using the 488-nm CW laser; the system is then aligned for other CW lasers automatically because they share the same
optical path.
Conjugate galvos GM-X, GM-Z, annulus mask and excitation objective ● TIMING 10–20 h
 CRITICAL The correct conjugation of the galvos to each other as well as the annulus mask AM and the rear pupil of the
excitation objective EO is critical to ensure that the beam only translates in x and z within the sample and does not also tilt
as the galvos are scanned. Each plane is conjugated to the next by using a pair of relay lenses in a 2f1 + 2f2 configuration.
2| With galvos GM-Z and GM-X powered at their mid-range control voltages, rotate GM-Z about its axis until the light beam
from GM-Z is centered on GM-X, and the light beam from GM-X is orthogonal to mask AM.
3| Conjugate GM-Z to GM-X by placing an inspection camera at the plane of GM-X, while modulating the angle of GM-Z at
a few Hertz. Iteratively adjust the axial positions of the two relay lenses until the mirror of GM-Z is in focus and does not
wobble when viewed through the inspection camera.
4| Place the inspection camera at the plane of mask AM, and repeat the procedure described in Step 3 until the mirror of
GM-X is in focus and is not wobbling at the plane of AM.
5| As a final check, place mask AM at this second conjugation plane, and center an annulus that will eventually produce
a Bessel beam of ~600 nm in diameter and 30 µm in length (e.g., NAOD = 0.4 and NAID = 0.36, where OD is outer diameter
and ID is inner diameter) in the sample on the impinging laser beam. Move the inspection camera to the side of AM opposite
the galvos, and then image the illuminated annulus onto the camera. Adjust the relay lenses if necessary to ensure that the
illumination through the annulus does not vary as the galvo tilts are modulated.
6| Conjugate annulus mask AM to excitation objective EO. The specimen chamber should be designed with a small beam
exit window opposite objective EO, along its axis. Fill the chamber with water, and view the annulus of light emerging from
the window on a white screen a minimum of 50 cm away. Place a mark on the screen at the point where the axis of EO
intersects the screen. Center the annular beam in the rear pupil of EO by translating the x and y right-angle prism mirrors
immediately before EO until the annular beam on the screen is centered with respect to the marked reference point. Adjust
the axial position of EO until the image of the annulus is in focus on the screen. This completes the initial centering and
conjugation of the beam relative to EO.
Align the detection path ● TIMING 10–20 h
7| Align the detection objective DO. Fill the specimen chamber with the prepared Alexa Fluor 488 hydrazide solution
and move mask AM laterally to the position of a round aperture (not an annulus) that produces a Gaussian beam of at
least 0.3 NA in the chamber. Set the output of the 488-nm laser to maximum, set galvos GM-Z and GM-X at their mid-range
positions, set the detection objective piezo to its mid-range position and remove detection tube lens TL1. Place a white
screen at a minimum of 50 cm from the rear pupil of objective DO and place a mark on the screen at the point where the
axis of DO intersects the screen. Adjust the axial position of DO until the beam of fluorescence emission emerging from
DO is collimated. Adjust the lateral position of DO until the beam emerging from DO is centered with respect to the marked
reference point.
8| Align the tube lens TL1. Remount tube lens TL1. Move mask AM laterally back to the position that produces a Bessel
beam of ~600 nm diameter and 30 µm length in the chamber. Adjust the axial position of TL1 until the image of the Bessel
beam is in focus on the detection camera. Adjust the lateral position of TL1 until the image of the beam is centered in the
field of view of the camera.
nature protocols | VOL.9 NO.5 | 2014 | 1093
protocol
z (µm)
30
10
z (µm)
z (µm)
z (µm)
z (µm)
z (µm)
x (µm)
x (µm)
x (µm)
z (µm)
z (µm)
z (µm)
z (µm)
© 2014 Nature America, Inc. All rights reserved.
z (µm)
z (µm)
x (µm)
–30
–30
–30
–30
Figure 5 | Microscope alignment. (a) An
a
b
c
d
unaligned Bessel beam tilted in the xy plane.
–15
–15
–15
–15
(b) An unaligned Bessel beam tilted in xy plane.
0
0
0
0
The right side of the beam is in focus, but the
left side is not. (c) A Bessel beam that, when
15
15
15
15
swept in the x direction, does not remain in
30
30
30
30
focus, so that the virtual light sheet and the
–30 –15 0
15 30
–30 –15 0
15 30
–30 –15 0
15 30
–30 –15 0
15
detection focal plane are not coincident.
y (µm)
y (µm)
y (µm)
y (µm)
(d) An aligned Bessel beam that is in focus at
–5
–5
–4
–4
all x positions across the scan plane. (e) The
e
f
g
h
–2
–2
cross-sectional intensity profile of a properly
0
0
aligned Bessel beam. (f) The cross-sectional
0
0
intensity profile of a Bessel beam caused by
2
2
nonuniform illumination across the annular
5
5
4
4
ring and mis-conjugation of the the annular
0
–5
0
5
–5
5
–2
2
–2
2
0
0
x (µm)
x (µm)
ring to the rear pupil of excitation objective EO.
x (µm)
x (µm)
(g) The detection wide-field xz PSF with
–10
–10
–10
–10
strong spherical aberration. (h) The detection
i
j
k
l
–5
–5
–5
–5
wide-field xz PSF after aberration correction
0
0
0
0
with the correction collar of DO. (i) xz MIP
view of a field of fluorescent beads imaged in
5
5
5
5
the sheet-scan mode showing symmetric PSFs
10
10
10
10
in the axial direction across the entire field of
–10 –5
0
5
10
–10 –5
0
5
10
–10 –5
0
5
10
–10 –5
0
5
view. (j) xz MIP view of a field of fluorescent
x (µm)
x (µm)
x (µm)
x (µm)
beads imaged in the sheet-scan mode, with a
–2
–2
–2
–2
1 µm offset between the center of the light
m
n
o
p
–1
–1
–1
–1
sheet and the focal plane of DO, showing an
identical axial asymmetry in the PSF across the
0
0
0
0
entire field of view. (k) Same as in i, except
1
1
1
1
acquired in the OS-SIM mode with N = 7 phases.
Note the substantially tighter PSF in the axial
2
2
2
2
–1
0
1
–1
0
1
–1
0
1
–1
0
1
direction. (l) Same as in j, except acquired
x (µm)
x (µm)
x (µm)
x (µm)
in the OS-SIM mode with N = 7 phases, and
a 0.3 µm offset between the excitation and detection planes. (m–p) Zoomed-in view of the same fluorescent bead marked in i–l. In general, the SIM
modes require an even tighter axial alignment than the swept-sheet modes.
9| Fine-tune the annular beam centration in the rear pupil of EO. With a Bessel beam excited in the AF 488 solution and
the galvos and piezo for DO all at mid-range, measure the width of the Bessel beam on the detection camera at two axial
locations symmetric about its center. Adjust the position of the annular beam entering objective EO along the z axis with
the appropriate right angle prism just before EO until the beam widths at both locations are the same (Fig. 5 shows an
unaligned case in which only one end of the beam is in focus). Adjust the position of the annular beam along the x axis
until the image of the Bessel beam is parallel to the x axis on the detection camera (Fig. 5a shows an unaligned case in
which the beam is not parallel to the x axis).
Fine-tuning ● TIMING 10–20 h
10| Fine-tune the conjugation to EO. Move the detection objective piezo to the top of the desired z-scan range.
Adjust the control voltage of galvo GM-Z until the center of the Bessel beam is in focus at this new position, as determined
by a minimum cross-sectional width midway along its length, as measured by the control utility. Repeat this process with
the piezo moved to the bottom of the desired z-scan range. Adjust the axial positions of the relay lenses between mask AM
and objective EO until the beam is in focus along its entire length at both the top and the bottom extremes of the z field
of view.
11| Calibrate the galvos. Move the detection objective piezo in 10-µm steps from −40 to 40 µm. At each position, adjust
the control voltage to galvo GM-Z until the Bessel beam is back in focus. Apply linear regression to the galvo voltage
versus the objective position measurements to determine the calibration constant of z beam displacement per volt applied
to the galvo. The correlation coefficient should be at least 0.999 for the Bessel beam to remain at the focal plane of
DO across the entire z field of view. Next, adjust the control voltage to galvo GM-X in approximately ten equal steps of
amplitude such that the beam moves across the entire x field of view as seen on the detection camera. Use the camera
images to determine the beam position for each control voltage and apply linear regression to determine the calibration
constant of x beam displacement per volt applied to the galvo.
? TROUBLESHOOTING
1094 | VOL.9 NO.5 | 2014 | nature protocols
protocol
12| Confirm conjugation and centration. Confirm that the light sheet swept by the Bessel beam remains in focus across
the entire xy field of view, for all z planes of the desired 3D stack (Fig. 5c shows a case where the beam is not in focus at
all x positions in a given plane, Fig. 5d shows the case of correct alignment). Repeat Steps 2–6 and 9 and 10 as needed to
achieve these conditions. This completes the alignment of the linear excitation beam.
© 2014 Nature America, Inc. All rights reserved.
13| Mount the rear pupil inspection camera. Mount the camera at the location specified in Figure 3a and, with the flip
mirror in the up position, adjust the camera position so that the annular light ring is in focus (i.e., conjugate to mask AM
and the rear pupil of objective EO) and centered in the field of view. Record and save the image of the annulus. Subsequent
alignment of the system, if necessary, can be accomplished by adjusting the position, focus and conjugation (i.e., no wobble
or intensity variation under scanning) of the excitation path as described in Steps 2–6 to match this stored image.
System characterization ● TIMING 10–20 h
14| Fine-tune and measure the wide-field detection PSF. Flush the Alexa Fluor 488 solution from the imaging chamber
with water until no fluorescence can be observed on the detection camera. Fill the imaging chamber with water. Place the
prepared fluorescent bead sample on the specimen holder. Translate the bead sample to the center of the imaging field.
Position a single isolated bead within the Bessel beam. Use the program utility to measure the 3D detection PSF. This works
by leaving the x galvo fixed, while moving the z galvo and detection objective piezo together to acquire a 3D image stack
of the bead. Check the xz view of this stack for any asymmetry in the axial PSF (Fig. 5g). Adjust the correction collar of DO
to achieve axial symmetry. If the range of the collar is insufficient, return the collar to its mid-range position, and iterate in
slight axial moves of objective DO and tube lens TL1 until axial symmetry is achieved (Fig. 5h). Check the xz and yz views of
the 3D stack for lateral asymmetry in the PSF. Iteratively adjust the lateral positions of objective DO and tube lens TL1 until
symmetry is achieved, while maintaining the image of the bead in approximately the same position on the camera. Record
the final, symmetric 3D detection PSF for eventual use in reconstruction of SR-SIM mode data, making sure that the signal
from only the desired target bead appears in the PSF data.
15| Characterize the Bessel beam cross-sectional intensity. Position an isolated fluorescent bead in focus at the common
center of the imaging field and the Bessel beam by using the sample translational stages. Use the program utility to measure
the cross-section of the Bessel beam. This works by using the galvos to raster-scan the excitation beam in 100-nm steps
across the bead over a ±5-µm range in the xz plane while recording an image at each position. The integrated fluorescence
recorded from the bead in each image gives a measure of the intensity of the beam at the associated xz position. The image
created by the entire xz set of these measurements yields the cross-section of the Bessel beam. The image should closely
approximate an ideal Bessel beam, with symmetric side lobes (Fig. 5e). If not (e.g., Fig. 5f), check with the rear pupil
inspection camera that the intensity is uniform across the annulus, and that both intensity and position of the annulus
remain constant as the galvos are scanned. Retune the conjugation of all elements and centration of the annular beam as
described in Steps 2–6 until a symmetric Bessel beam is achieved.
? TROUBLESHOOTING
16| Characterize the Bessel swept-sheet PSF. Position a field of fluorescent beads within the volume to be imaged. Take a
3D image stack of this field in the swept-sheet mode across an image volume of ~50 × 50 × 50 µm. The 3D image of each
fluorescent bead within the field corresponds to the Bessel sheet-scan PSF at the bead position. All PSFs should be identical
and symmetric in the axial direction across the entire imaging field (Fig. 5i), although they may be less intense at the edges
of the field in the y direction owing to finite extent of the beam. Adjust the initial offset of the detection objective piezo to
achieve symmetry if the PSFs are uniformly asymmetric in the same axial direction. Return to Steps 10–12 if the PSFs are not
uniform or are asymmetric in different directions. Record the image of an isolated bead taken in the swept sheet mode for
eventual use as the reference 3D PSF in deconvolution of image data.
? TROUBLESHOOTING
17| Characterize the Bessel OS-SIM PSF. Take a 3D image stack of a field of fluorescent beads with the Bessel OS-SIM mode
with a periodic pattern having H = 3 or 4 harmonics, requiring N = 7 or 9 images per plane. Reconstruct the final 3D stack
from the raw data by using the OS-SIM algorithm. The 3D image of each fluorescent bead within the field corresponds to the
Bessel OS-SIM PSF at the bead position. If the xz PSFs are uniformly asymmetric axially (Fig. 5l), adjust the offset position
of the detection objective piezo until symmetry is achieved (Fig. 5k). Return to Steps 10–12 if the PSFs are not uniform or
asymmetric in different directions. The axial resolution of the reconstructed OS-SIM image (Fig. 5k) should be substantially
better than that taken by the Bessel swept-sheet mode (Fig. 5k). Record the image of an isolated bead taken in the OS-SIM
mode for eventual use as the reference 3D PSF in deconvolution of image data. This completes the system alignment for
linear excitation with a single Bessel beam.
? TROUBLESHOOTING
nature protocols | VOL.9 NO.5 | 2014 | 1095
protocol
(Optional) Bessel beam array alignment ● TIMING 1–2 h
18| Create an array of parallel Bessel beams. Parallel Bessel beams can be used, if desired, to increase the speed of Bessel
plane SIM and/or to reduce phototoxicity by reducing the peak intensity within the sample. Place diffractive optical element
DOE against the annulus mask on the side facing the galvos to create multiple excitation beams simultaneously.
? TROUBLESHOOTING
19| Check the DOE alignment and beam spacing. Fill the imaging chamber with Alexa Fluor 488 dye solution. Rotate the
DOE around its axis until all of the parallel excitation beams are in focus simultaneously (Fig. 5d). Check that the fan angle
of the DOE is large enough so that the side lobes of adjacent beams do not interfere. Return to Step 12 and repeat other
alignment steps as needed.
© 2014 Nature America, Inc. All rights reserved.
20| Select the beam spacing for Bessel SIM modes. Use the image of the parallel beams in dye solution to measure the exact
spacing between beams. For the OS-SIM and SR-SIM modes, make sure that period T of the written SIM grating pattern is
given by the beam spacing divided by an integer.
(Optional; for TP Bessel beam plane illumination and TP Bessel SIM mode) TP mode alignment ● TIMING 5–10 h
21| Create an excitation annulus for TP imaging. Adjust the Chameleon laser source to a visible wavelength (e.g., 770 nm).
Attenuate the beam power by using a half-wave plate and a polarizer. Align the Pockels cell according to the Pockels cell
alignment manual. Collimate and expand the NIR beam to an ~10-mm beam diameter. Place another inspection camera at the
focal plane FP1 (Fig. 3b) of lens L7. An annular focus created by axicon AX and lens L7 should be visible. Adjust the input
beam centration and perpendicularity to AX until the intensity is uniform across the annular beam.
22| Perform initial conjugation and alignment of the annular beam. Remove this secondary inspection camera. Insert lens
L9 into the optical path and conjugate the annular focus first to galvo mirror GM-Z. Tune the Chameleon laser to the desired
NIR wavelength for TP excitation. Adjust the axial positions of relay lenses L8 and L9 until the annular beam is in focus at
galvo GM-Z. Adjust the lateral positions of L8 and L9 as well as the tilt angle of dichroic mirror DC1 until the annular beam
passes through the desired annulus in mask AM. View the filtered annular beam on the rear pupil inspection camera from
Step 13. Again adjust L8, L9 and DC1 until the intensity of the annular beam is uniform.
? TROUBLESHOOTING
23| Fine-tune conjugation and alignment. The differences in the focal lengths at visible and NIR wavelengths of all
relay lenses in the optical path may cause the GM-Z, GM-X, AM and the rear pupil of objective EO to no longer be mutually
conjugated. Test the conjugation of these elements and centration to the rear pupil of EO by using the same procedures as
those given above for linear excitation. Confirm with an isolated fluorescent bead the uniformity of the scanned TP light
sheet. Confirm with a field of beads the uniformity, symmetry and quality of the overall 3D PSF in the TP swept-sheet
and SIM modes. Record the image of an isolated bead taken in the TP swept-sheet mode for eventual use as the reference
3D PSF in deconvolution of image data.
Image acquisition ● TIMING 0.5–1 h
 CRITICAL We have provided guidelines below for a typical experiment. We do not describe sample preparation, as this
will be the same as for conventional wide-field and confocal fluorescence microscopy. The sample mounting method will be
dictated by the instrument design. Examples of sample preparation and mounting can be found in previous publications10,17.
24| Estimate the desired excitation beam length and thickness on the basis of the sample size, desired spatial resolution,
acquisition mode (objective scan or sample scan), imaging speed, number of time points and sample brightness.
Use Table 2 to select an appropriate annulus on the mask on the basis of this estimation. Select the beam that is just
long enough to cover the imaging field of interest and thin enough to give the desired spatial resolution in order to
minimize the influence of the Bessel beam side lobe.
25| Fill the imaging chamber with the corresponding dye solution. Verify the Bessel beam alignment as in Step 9.
26| Fill the sample chamber with the desired imaging buffer. Characterize the system PSF again as described in Steps 15–17.
27| The system is ready for room temperature imaging. If room temperature is compatible with the sample, load the
sample and then proceed to Step 32 below. For imaging at elevated temperature (e.g., 37 °C), first perform Steps 28–32.
? TROUBLESHOOTING
1096 | VOL.9 NO.5 | 2014 | nature protocols
protocol
Table 2 | Approximate Bessel beam length and FWHM of the central peak for selected annuli with different illumination NAOD and NAID values.
NAID
NAOD
Col 1
Col 2
Col 3
Col 4
Col 5
Col 6
Col 7
Col 8
Row 1
0.64
0.630
0.623
0.613
0.600
0.585
0.560
0.520
0.450
Row 2
0.60
0.588
0.582
0.572
0.555
0.540
0.505
0.470
0.400
Row 3
0.55
0.537
0.530
0.517
0.500
0.480
0.440
0.390
0.300
Row 4
0.50
0.485
0.477
0.465
0442
0.420
0.375
0.325
0.210
Row 5
0.40
0.391
0.381
0.371
0.350
0.325
0.295
0.225
0
Row 6
0.30
0.290
0.280
0.275
0.260
0.233
0.190
0.115
0
Row 7
0.20
0.195
0.190
0.185
0.180
0.170
0.160
0.130
0
© 2014 Nature America, Inc. All rights reserved.
Approximate beam FWHM length (mm)
Row 1
75
50
30
20
15
10
7.5
5
Row 2
75
50
30
20
15
10
7.5
5
Row 3
75
50
30
20
15
10
7.5
5
Row 4
75
50
30
20
15
10
7.5
5
Row 5
150
75
50
30
20
15
10
7
Row 6
200
100
75
50
30
20
15
12.5
Row 7
>300
300
200
150
100
75
50
30
Approximate central peak FWHM width (mm)
Row 1
0.28
0.28
0.28
0.28
0.28
0.30
0.30
0.32
Row 2
0.30
0.30
0.30
0.30
0.30
0.32
0.32
0.34
Row 3
0.32
0.32
0.34
0.34
0.34
0.36
0.38
0.4
Row 4
0.36
0.36
0.36
0.38
0.38
0.40
0.42
0.46
Row 5
0.46
0.46
0.46
0.48
0.48
0.50
0.54
0.62
Row 6
0.6
0.62
0.62
0.62
0.66
0.70
0.78
0.84
Row 7
0.92
0.92
0.92
0.92
0.96
0.96
0.11
0.13
28| Use a water bath to heat a reservoir of water. Turn on the water pump to circulate the heated water through the
channels in the walls of the sample chamber.
29| Place a sample with a field of fluorescent beads in the chamber. Monitor the temperature of the imaging buffer in the
chamber and periodically image the field of beads in the swept-sheet or OS-SIM modes as the temperature rises. Adjust the
initial offset of the detection objective piezo and the correction collar of DO to maintain a symmetric, aberration-free axial
PSF from the beads.
30| Adjust the temperature of the water bath as needed and repeat Step 29 periodically until the buffer in the chamber
equilibrates at the desired imaging temperature.
31| Replace the imaging buffer in the chamber with fresh buffer, prewarmed to the desired equilibrium temperature if you
are not imaging at room temperature. The system is now ready for imaging.
nature protocols | VOL.9 NO.5 | 2014 | 1097
protocol
© 2014 Nature America, Inc. All rights reserved.
32| During imaging, it is crucial to keep the excitation light sheet centered on the focal plane of objective DO to achieve
the desired performance, especially for Bessel SIM. Both of the following two methods can be used to ensure that the
excitation light sheet remains in focus.
(A) Maximize the in-focus modulation depth
(i) Occasionally apply a fixed, periodic modulation pattern, such as that required for SIM, and adjust the offset of the
detection objective piezo until the maximum modulation depth is observed.
(B) Maximize the SNR of OS-SIM reconstructed results
(i) Occasionally acquire an image of the sample in OS-SIM mode and adjust the piezo offset until the SNR of the
OS-SIM reconstruction is a maximum. The OS-SIM reconstruction depends only on the fundamental modulation
harmonic of the illumination pattern whose amplitude is maximum when the modulation pattern is in focus.
? TROUBLESHOOTING
Data processing ● TIMING variable; ~1–3 min per image volume
33| Analyze the data by using either option A, B or C, depending on how the data were acquired. Note that for option C,
the raw 3D data stack with N images per plane of N different phases of the applied grating are analyzed by using the
algorithm originally developed for wide-field 3D-SIM. A thorough description with more details is available elsewhere14.
(A) Data obtained in linear or TP swept-sheet modes with an objective scan
(i) Measure the system PSF by using the corresponding imaging method as described in Step 16. The PSF must be
measured on a single, isolated fluorescent bead at voxel sizes that are smaller in all three dimensions than half
the theoretical resolution limit.
(ii) Subtract the noise background from the measured PSF data volume.
(iii) Deconvolve the acquired data by using the Richardson-Lucy 3D deconvolution algorithm in MATLAB and the
processed PSF.
(B) Data obtained in linear or TP swept-sheet modes with a sample scan
(i) Measure the system PSF as described in Step 33A(i).
(ii) Process the measured PSF as described in Step 33A(ii).
(iii) Transform the data from the skewed scanning coordinates back to the optical xyz coordinates.
(iv) Deconvolve the transformed data as described in Step 33A(iii).
(C) Data obtained in SR-SIM mode
(i) Pre-process the original images by using flat-field correction and soften the image edges.
(ii) Separate the spectral information components. As discussed above, the raw data consist of three or more different
information components that are originally located at different positions in frequency space, and need to be separated
first. The raw data, including a total of 2N + 1 images per image plane, can be formulated as a linear combination of the
information components in frequency space and information components need to be calculated from the raw images.
(iii) Measure the system PSF to acquire the effective OTF of each information component, Om. This is done by acquiring
a Bessel SIM data set on a single fluorescent bead under exactly the same experimental conditions as the data to
be reconstructed. The voxel size of both the PSF and the raw data must be smaller than the Nyquist limit.
(iv) Apply an inverse filter and reassemble each information component back to its appropriate location in frequency
space. Transform back to real space to produce the final reconstructed image with extended resolution.
? TROUBLESHOOTING
Troubleshooting advice can be found in Table 3.
Table 3 | Troubleshooting table.
Step
Problem
Possible reason
Solution
11
The calibration curve is not
linear
The silicone plastic film is too tight,
which induces high resistance to the
DO scanning piezo
Verify that the silicone plastic film does not resist the
DO movement, so that the DO can be scanned freely in
the axial direction
15
The wide-field PSF is not
ideal; typical aberrations are
spherical aberration, coma
and astigma
The detection path is not properly
aligned. The optical axis of the TL
and the DO are not coaxial
Repeat Steps 7 and 8 until ideal wide-field PSF is
obtained. Adjust the axial positions of DO and TL to
correct spherical aberration. Use an objective collar
ring if available. Adjust the lateral positions of TL and
DO to correct coma and astigma
(continued)
1098 | VOL.9 NO.5 | 2014 | nature protocols
protocol
© 2014 Nature America, Inc. All rights reserved.
Table 3 | Troubleshooting table (continued).
Step
Problem
Possible reason
Solution
16
The Bessel beam intensity
profile is not ideal
The conjugation between two galvos,
the AM mask and the excitation
objective is not properly aligned
The annulus intensity is not uniform
Confirm the conjugation is correct. Repeat Steps 2–6
if necessary
Use the conjugation camera to check the
illumination annulus intensity. Adjust the laser or the
annulus mask lateral position to make it uniform
17, 18
The PSFs are not symmetrical
in the axial direction
The illumination light sheet is not in
focus
Adjust the offset of the detection objective piezo until
the PSF is symmetrical in the axial direction
22
The intensity of the annular
focus is not uniform
The axicon AX, the lens L7 and the
incident NIR beam are not coaxial
The incident beam is not
perpendicular to the axicon AX
Adjust the lateral positions of the NIR beam, AX and
L7 until the annular focus intensity is uniform
27
Spherical aberration is
observed after changing the
imaging buffer
The imaging buffer refraction index
is different from that of water
Use the DO collar ring TL to correct spherical
aberration. Adjust the axial positions of DO and TL if
there is no DO collar ring available
32
The DO focus keeps drifting
during imaging
The imaging chamber temperature
is not stable, which causes the
DO focus to drift
Give the heating system enough time to let the
imaging chamber temperature stabilize
Modulation patterns cannot
be observed during image
acquisition
The illumination light sheet is
not in focus
Find the focus as discussed in Step 32
Poor SNR with the acquired
images
The modulation pattern period is
too small
Increase the modulation pattern period and the
corresponding number of raw images
The laser intensity is not high enough
Increase the excitation laser intensity
The selected Bessel beam’s NA is
too high
Select a low-NA Bessel beam
● TIMING
Step 1, collimate and combine CW laser beams: ~10 h
Steps 2–6, conjugate galvos GM-X, GM-Z, annulus mask and excitation objective: 10–20 h
Steps 7–9, align the detection path: 10–20 h
Steps 10–13, fine-tuning: 10–20 h
Steps 14–17, system characterization: 10–20 h
Steps 18–20, (optional) Bessel beam array alignment: 1–2 h
Steps 21–23, (optional) TP mode alignment: 5–10 h (Steps 21–23 only need to be performed for TP Bessel beam plane
illumination and TP Bessel SIM mode)
Steps 24–32, image acquisition: 0.5–1 h
Step 33, data processing: variable; ~1–3 min per image volume, depending on the image data size and the computation
power of the data-processing computer
ANTICIPATED RESULTS
Figure 6 shows the xy and xz maximum intensity projection (MIP) views of fluorescently labeled microtubules in a fixed
LLC-PK1 cell, imaged with the SR-SIM mode in PBS. An annular illumination pattern of NAOD = 0.59, NAID = 0.55 was used,
yielding a Bessel beam of ~410 nm in central peak thickness and ~15 µm in length. This represents a relatively high NA yet a
short beam for Bessel beam plane illumination. Both of these attributes are desirable, as long as they are consistent with the
requirements of the sample. High NA yields high spatial frequencies for higher resolution SIM reconstruction and short beam
length results in low side lobe energy and thus less out-of-focus background, which can otherwise compromise the SNR in
SIM reconstructions. For the case of the thin cultured cell in Figure 6a,b, these parameters work well because a 15-µm beam
is sufficient to cover the ~10-µm thickness of the cell in sample-scan mode, and because the cell is so thin that even at high
NA the sample-induced aberrations are negligible.
nature protocols | VOL.9 NO.5 | 2014 | 1099
© 2014 Nature America, Inc. All rights reserved.
protocol
Figure 6 | Examples of Bessel beam structured plane illumination.
(a,b) xy and xz MIP views of a fixed LLC-PK1 cell with fluorescently
labeled microtubules imaged in the SR-SIM mode by using nine phases
and comb modulation. (c,d) Magnified views of the boxed regions in a
and b. (e) Line width measurements from the microtubules c and d in the
x, y and z directions. (f) A raw image during the acquisition with comb
modulation, with arrows showing the maxima of the grating pattern within
the sample. (g) Magnified view of the boxed region in f. (h,i) xy and xz MIP
views of a Dictyostelium discoideum cell imaged in the SR-SIM mode with
sinusoidal modulation. (j) A raw image frame during the acquisition, with
arrows showing the very weak modulation pattern produced by sinusoidal
modulation. (k) MIP view of the real-time, in-situ OS-SIM reconstruction from
the sine mode data that can be used to indicate whether the modulation
amplitude and excitation power are sufficient to obtain a good SR-SIM
reconstruction. (l,m) xy and xz MIP views of a C. elegans embryo imaged by
using the SR-SIM mode with five phases and comb modulation. (n) A raw
image ~15 µm below the sample surface, with arrows showing the maxima
of the comb pattern. Note that the pattern is stronger on the left side
than on the right, as optical aberrations and scattering cause distortions
in the pattern as the beam prorogates from left to right. (o) xy MIP of the
real-time, in-situ OS-SIM reconstruction used for visual feedback during
acquisition. Note that the OS-SIM reconstruction has both lower x resolution
and lower SNR than the final SR-SIM reconstruction in l. Scale bars, 5 µm
(a,b,f,h,l), 1 µm (c,d,g).
a
b
c
f
d
x
e
0
x
y
x linewidth
y linewidth
0.2 0.4 0.6 0.8 0
h
x
l
x
g
z linewidth
0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 1.0
i
y
z
x
k
n
o
z
m
x
j
y
z
A comb pattern of 1.8 µm period, visible in the raw data
frame in Figure 6f,g, was used to create the grating
pattern for SIM. At NAOD = 0.59, this required acquisition
of N = 9 phase-stepped images per plane for the 3D SIM reconstruction. The entire raw data stack is shown in
Supplementary Video 1. A large number of phases N improves the SNR of the final reconstruction, because it makes use
of more information and includes low frequency harmonics of higher amplitude. It does, however, take longer to acquire and
expends more of the limited photobleaching-dictated photon budget. Fortunately, for fixed samples such as those shown
here, these are not important concerns.
With these imaging conditions, the theoretical spatial resolution in Figure 6a,b is ~160 nm in the x direction, ~240 nm
in y and ~290 nm in z. These predictions are consistent with the experimental results (Fig. 6e).
Figure 6h,i shows MIP views of a living Dictyostelium discoideum cell expressing F-actin (labeled by GFP), imaged by using
the SR-SIM mode with sinusoidal modulation. This modulation pattern was chosen because it requires the minimum N = 3
images per plane required for SIM reconstruction, and hence is more compatible with rapid and noninvasive imaging of
quickly moving, light-sensitive specimens such as these. An illumination annulus of NAOD = 0.43, NAID = 0.4 was chosen
to create a longer Bessel beam in order to cover the potential migration of the cell over the course of the imaging.
A comparatively long modulation period of 3.5 µm was used to improve the optical sectioning capability. Although this
still results in near-optimal axial resolution improvement to ~350 nm in the SIM reconstruction, the lateral resolution
improvement beyond the ~240-nm limit of diffraction is negligible.
One limitation of using sinusoidal modulation for Bessel plane SR-SIM is that the ratio of the amplitude of the
fundamental harmonic to the DC harmonic is typically much smaller (e.g., Fig. 6j and Supplementary Video 2) than that
with the comb function, thus requiring a higher SNR in the raw data to get a satisfactory reconstruction. As a quick check
to see whether the SNR of the raw data is sufficient, a trial 3D stack can be collected and analyzed in situ with the OS-SIM
algorithm (Fig. 6k) and the power adjusted accordingly just before the desired 4D data set is collected.
Multicellular specimens typically require longer and thicker Bessel beams, both to span the complete specimen and
to reduce the sensitivity to sample-induced aberrations, which are a much greater concern in such specimens.
For example, the live C. elegans embryo in Figure 6m was imaged with a five-phase Bessel plane SR-SIM with comb
modulation at NAOD = 0.31 and NAID = 0.27. This produced a Bessel beam ~40 µm long, which was sufficient to image
the ~60 × 35 µm embryo, provided that the short axis of the specimen was aligned parallel to the beam. In general,
all asymmetric specimens should be mounted with their short axes parallel to the beam, to permit the shortest possible
beam to be used with the least possible amount of energy in the side lobes. To cover the scan of the long axis more quickly
and to reduce photodamage, a DOE was added producing seven parallel Bessel beams. A SIM pattern period of 2.12 µm
was chosen to ensure the strength of the fundamental modulation order. The theoretical spatial resolution is ~200 nm in
the x direction, ~240 nm in y and ~430 nm in z. As above, the OS-SIM algorithm can be used (Fig. 6o) to verify that
modulation amplitude and excitation power are sufficient to yield good-quality SR-SIM reconstructions; it can be very
difficult to know this from the raw images directly (Fig. 6n and Supplementary Video 3).
1100 | VOL.9 NO.5 | 2014 | nature protocols
protocol
Note: Any Supplementary Information and Source Data files are available in the
online version of the paper.
Acknowledgments We thank T. Planchon for help with the instrument;
H. White and S. Michale for cell culture, transfection and staining; and we thank
D. Milkie of Coleman Technologies for custom instrument control software.
L.G. acknowledges the support of a Stony Brook University start-up grant.
AUTHOR CONTRIBUTIONS L.G. and E.B. conceived the project and designed the
experiments; L.G. and B.-C.C. performed experiments; L.G. and L.S. analyzed the
data; and L.G. and E.B. wrote the paper with the input of all other authors.
COMPETING FINANCIAL INTERESTS The authors declare competing financial
interests: details are available in the online version of the paper.
© 2014 Nature America, Inc. All rights reserved.
Reprints and permissions information is available online at http://www.nature.
com/reprints/index.html.
1. Siedentopf, H. & Zsigmondy, R. Über Sichtbarmachung und
Groessenbestimmung ultramikroskopischer Teilchen, mit besonderer
Anwendung auf Goldrubinglaesern. Ann. d. Physik 10, 1–39 (1903).
2. Huisken, J., Swoger, J., Del Bene, F., Wittbrodt, J. & Stelzer, E.H.K.
Optical sectioning deep inside live embryos by selective plane illumination
microscopy. Science 305, 1007–1009 (2004).
3. Greger, K., Swoger, J. & Stelzer, E.H.K. Basic building units and properties
of a fluorescence single plane illumination microscope. Rev. Sci. Instrum.
78, 023705 (2007).
4. Keller, P.J., Schmidt, A.D., Wittbrodt, J. & Stelzer, E.H.K. Reconstruction
of zebrafish early embryonic development by scanned light-sheet
microscopy. Science 322, 1065–1069 (2008).
5. Tomer, R., Khairy, K., Amat, F. & Keller, P.J. Quantitative high-speed
imaging of entire developing embryos with simultaneous multiview
light-sheet microscopy. Nat. Methods 9, 755–763 (2012).
6. Krzic, U., Gunther, S., Saunders, T.E., Streichan, S.J. & Hufnagel, L.
Multiview light-sheet microscope for rapid in toto imaging. Nat. Methods
9, 730–733 (2012).
7. Ahrens, M.B., Orger, M.B., Robson, D.N., Li, J.M. & Keller, P.J. Whole-brain
functional imaging at cellular resolution using light-sheet microscopy.
Nat. Methods 10, 413–420 (2013).
8. Keller, P.J. Imaging morphogenesis: technological advances and
biological insights. Science 340, 1234168 (2013).
9. Keller, P.J. et al. Fast, high-contrast imaging of animal development
with scanned light-sheet–based structured-illumination microscopy.
Nat. Methods 7, 637–642 (2010).
10. Planchon, T.A. et al. Rapid three-dimensional isotropic imaging of living
cells using Bessel beam plane illumination. Nat. Methods 8, 417–423
(2011).
11. Durnin, J., Miceli, J.J. & Eberly, J.H. Comparison of Bessel and Gaussian
beams. Opt. Lett. 13, 79–80 (1988).
12. Truong, T.V., Supatto, W., Koos, D.S., Choi, J.M. & Fraser, S.E. Deep and
fast live imaging with two-photon scanned light-sheet microscopy.
Nat. Methods 8, 757–760 (2011).
13. Neil, M.A.A., Juškaitis, R. & Wilson, T. Method of obtaining optical
sectioning by using structured light in a conventional microscope.
Opt. Lett. 22, 1905–1907 (1997).
14. Gustafsson, M.G. Surpassing the lateral resolution limit by a factor
of two using structured illumination microscopy. J. Microsc. 198, 82–87
(2000).
15. Gustafsson, M.G.L. et al. Three-dimensional resolution doubling in
wide-field fluorescence microscopy by structured illumination.
Biophys. J. 94, 4957–4970 (2008).
16. Shao, L., Kner, P., Rego, E.H. & Gustafsson, M.G.L. Super-resolution 3D
microscopy of live whole cells using structured illumination. Nat. Methods
8, 1044–1046 (2011).
17. Gao, L. et al. Noninvasive imaging beyond the diffraction limit of 3D
dynamics in thickly fluorescent specimens. Cell 151, 1370–1385
(2012).
18. Roh-Johnson, M. et al. Triggering a cell shape change by exploiting
preexisting actomyosin contractions. Science 335, 1232–1235 (2012).
19. Habib, S.J. et al. A localized Wnt signal orients asymmetric stem cell
division in vitro. Science 339, 1445–1448 (2013).
20. Majoul, I.V. et al. Fast structural responses of gap junction membrane
domains to AB5 toxins. Proc. Natl. Acad. Sci. USA 110, E4125–4133
(2013).
21. Schermelleh, L., Heintzmann, R. & Leonhardt, H. A guide to
super-resolution fluorescence microscopy. J. Cell Biol. 190, 165–175
(2010).
22. Cella Zanacchi, F. et al. Live-cell 3D super-resolution imaging in thick
biological samples. Nat. Methods 8, 1047–1049 (2011).
23. Cella Zanacchi, F., Lavagnino, Z., Faretta, M., Furia, L. & Diaspro, A.
Light-sheet confined super-resolution using two-photon photoactivation.
PLoS ONE 8, e67667 (2013).
24. Friedrich, M., Gan, Q., Ermolayev, V. & Harms, G.S. STED-SPIM: stimulated
emission depletion improves sheet illumination microscopy resolution.
Biophys. J. 100, L43–L45 (2011).
25. Fahrbach, F.O., Simon, P. & Rohrbach, A. Microscopy with
self-reconstruction beams. Nat. Photon. 4, 780–785 (2010).
26. Bélanger, P.A. & Rioux, M. Ring pattern of a lens-axicon doublet
illuminated by a Gaussian beam. Appl. Opt. 17, 1080–1088 (1978).
nature protocols | VOL.9 NO.5 | 2014 | 1101